Adaptive Sample-level Graph Combination for Partial Multiview Clustering.

Area of Science:

• Machine Learning
• Data Mining
• Computer Science

Background:

• Multiview clustering assumes complete data, but real-world data often has missing information (partial multiview clustering).
• Existing partial multiview clustering methods often treat all data views equally, ignoring their varying importance.
• Missing data in multiview clustering significantly degrades performance.

Purpose of the Study:

• To propose a novel adaptive method for partial multiview clustering that automatically adjusts the contributions of different views.
• To address the challenge of missing data in multiview clustering by leveraging complete and incomplete samples.
• To improve the performance of clustering algorithms when dealing with incomplete datasets.

Main Methods:

• Developed an adaptive method for partial multiview clustering.
• Divided samples into complete and incomplete sets.
• Established a joint learning mechanism with a two-term optimization model to integrate complementary information and adaptively update view importance.

Main Results:

• The proposed method effectively handles missing data in multiview clustering.
• Experimental results on real-world datasets demonstrate the method's effectiveness and efficiency.
• The adaptive weighting of views significantly enhances clustering performance.

Conclusions:

• The novel adaptive method improves partial multiview clustering by dynamically adjusting view contributions.
• The joint learning mechanism successfully connects complete and incomplete samples for better clustering.
• This approach offers a robust solution for multiview clustering with missing data.